topic log likelihood of the VECM in SAS Programming

log likelihood of the VECM

Golf — Tue, 31 Mar 2020 06:50:33 GMT

Hello, I use the following proc to do cointegration test.

proc varmax data=Example_var;
model LM1 LIP TB3/ p=2 cointtest=(johansen=(normalize=LM1)) ;
run;

I got the following results:

Can anyone inform me how to estimate the log-likelihood? (I try the same data set in another software, but the log-likelihood is different. I'd like to know which one is the correct one.)

Thank You.

Re: log likelihood of the VECM

Rick_SAS — Tue, 07 Apr 2020 11:39:40 GMT

The definition of the log-likelihood that PROC VARMAX uses is shown in the SAS documentation for PROC VARMAX

It is sometimes difficult to compare LL across software products because you can drop constant terms when you optimize the LL. So if LL(x) = A + B*f(x), some software might use and report f(x) even though it isn't the full LL. Also, MLE methods depend on the optimization algorithms, convergence criteria, and initial guesses for the solution.